While working on a performance issue, I have seen that many a
times we have Long IO Errors on Tempdb database, which indicates that we have
scope of optimization for our tempdb Database.
As per Microsoft, there are a few Recommendation which we
should Test on Benchmark Servers for sure before making any changes in
production environment.
We could see the below options for Tempdb Optimization. These
are always the best practices which often increase performance. As we know the
Role of Tempdb has increased Drastically after the introduction of Sql Server
2005 so , why not Try them on Bench and implement on production.
To ensure efficient tempdb operation:
· Create one tempdb file per physical CPU
core. (But should ideally be used when we
have data files on separate disk spindles).
This reduces page free space (PFS) contention.
This reduces page free space (PFS) contention.
· Pre-size the tempdb files, and make the
files equal in size.
· Do not rely on autogrow.
· Use startup trace flag 1118.
For more information about this SQL Server trace flag, pls. go through this link http://support.microsoft.com/kb/328551
Note: There’s only one tempdb per instance, and lots of things use it, so
it’s often a performance bottleneck. You guys know that already. But when does
a performance problem merit creating extra tempdb data files?
There should be only 8 data files even we have more than 8 processor as
per Microsoft recommendation unless there is in-memory contention
Hope it will help you.
Brgds,
Chhavinath Mishra
Sr. Database
Administrator
Microsoft Certified IT Professional (MCITP)
No comments:
Post a Comment